README.md

Concrete Quantizer

A Rust implementation of a quantizer compatible with Concrete ML's quantization scheme.

Overview

This project provides a Rust-based quantizer that can load quantization parameters generated by Concrete ML's Python implementation. It allows for quantization and dequantization of values using the same logic as the Concrete ML python version.

Features

• Load quantizer parameters from JSON files generated by Concrete ML
• Quantize floating-point values to integers
• Dequantize integer values back to floating-point
• Compatibility testing with Python-generated results

Development Setup

Python Dependencies

To generate test data, you'll need Python with Concrete ML:

pip install -r requirements.txt

Then generate the test data for the logistic regression quantizer:

python scripts/generate_logistic_regression_quantizer_params.py